journal: properly implement matching with multiple matches
This commit is contained in:
parent
1cc101f1e5
commit
de7b95cdc3
|
@ -65,6 +65,7 @@ _packed_ struct DataObject {
|
|||
|
||||
_packed_ struct EntryItem {
|
||||
uint64_t object_offset;
|
||||
uint64_t hash;
|
||||
uint64_t prev_entry_offset;
|
||||
uint64_t next_entry_offset;
|
||||
};
|
||||
|
|
|
@ -805,6 +805,7 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
|
|||
|
||||
xor_hash ^= le64toh(o->data.hash);
|
||||
items[i].object_offset = htole64(p);
|
||||
items[i].hash = o->data.hash;
|
||||
}
|
||||
|
||||
r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqno, ret, offset);
|
||||
|
|
|
@ -31,7 +31,7 @@
|
|||
#include "log.h"
|
||||
|
||||
int main(int argc, char *argv[]) {
|
||||
int r;
|
||||
int r, i;
|
||||
sd_journal *j = NULL;
|
||||
|
||||
log_set_max_level(LOG_DEBUG);
|
||||
|
@ -46,6 +46,14 @@ int main(int argc, char *argv[]) {
|
|||
goto finish;
|
||||
}
|
||||
|
||||
for (i = 1; i < argc; i++) {
|
||||
r = sd_journal_add_match(j, argv[i], strlen(argv[i]));
|
||||
if (r < 0) {
|
||||
log_error("Failed to add match: %s", strerror(-r));
|
||||
goto finish;
|
||||
}
|
||||
}
|
||||
|
||||
SD_JOURNAL_FOREACH(j) {
|
||||
const void *data;
|
||||
size_t length;
|
||||
|
|
|
@ -28,13 +28,14 @@
|
|||
#include "journal-file.h"
|
||||
#include "hashmap.h"
|
||||
#include "list.h"
|
||||
#include "lookup3.h"
|
||||
|
||||
typedef struct Match Match;
|
||||
|
||||
struct Match {
|
||||
char *data;
|
||||
size_t size;
|
||||
uint64_t hash;
|
||||
uint64_t le_hash;
|
||||
|
||||
LIST_FIELDS(Match, matches);
|
||||
};
|
||||
|
@ -46,6 +47,7 @@ struct sd_journal {
|
|||
uint64_t current_field;
|
||||
|
||||
LIST_HEAD(Match, matches);
|
||||
unsigned n_matches;
|
||||
};
|
||||
|
||||
int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
|
||||
|
@ -71,8 +73,11 @@ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
|
|||
}
|
||||
|
||||
memcpy(m->data, data, size);
|
||||
m->le_hash = hash64(m->data, size);
|
||||
|
||||
LIST_PREPEND(Match, matches, j->matches, m);
|
||||
j->n_matches ++;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
@ -86,6 +91,8 @@ void sd_journal_flush_matches(sd_journal *j) {
|
|||
free(m->data);
|
||||
free(m);
|
||||
}
|
||||
|
||||
j->n_matches = 0;
|
||||
}
|
||||
|
||||
static int compare_order(JournalFile *af, Object *ao, uint64_t ap,
|
||||
|
@ -155,6 +162,103 @@ static int compare_order(JournalFile *af, Object *ao, uint64_t ap,
|
|||
return 0;
|
||||
}
|
||||
|
||||
static int move_to_next_with_matches(sd_journal *j, JournalFile *f, Object **o, uint64_t *p) {
|
||||
int r;
|
||||
uint64_t cp;
|
||||
Object *c;
|
||||
|
||||
assert(j);
|
||||
assert(f);
|
||||
assert(o);
|
||||
assert(p);
|
||||
|
||||
if (!j->matches) {
|
||||
/* No matches is easy, just go on to the next entry */
|
||||
|
||||
if (f->current_offset > 0) {
|
||||
r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &c);
|
||||
if (r < 0)
|
||||
return r;
|
||||
} else
|
||||
c = NULL;
|
||||
|
||||
return journal_file_next_entry(f, c, o, p);
|
||||
}
|
||||
|
||||
/* So there are matches we have to adhere to, let's find the
|
||||
* first entry that matches all of them */
|
||||
|
||||
if (f->current_offset > 0)
|
||||
cp = f->current_offset;
|
||||
else {
|
||||
r = journal_file_find_first_entry(f, j->matches->data, j->matches->size, &c, &cp);
|
||||
if (r <= 0)
|
||||
return r;
|
||||
|
||||
/* We can shortcut this if there's only one match */
|
||||
if (j->n_matches == 1) {
|
||||
*o = c;
|
||||
*p = cp;
|
||||
return r;
|
||||
}
|
||||
}
|
||||
|
||||
for (;;) {
|
||||
uint64_t np, n;
|
||||
bool found;
|
||||
Match *m;
|
||||
|
||||
r = journal_file_move_to_object(f, cp, OBJECT_ENTRY, &c);
|
||||
if (r < 0)
|
||||
return r;
|
||||
|
||||
n = journal_file_entry_n_items(c);
|
||||
|
||||
/* Make sure we don't match the entry we are starting
|
||||
* from. */
|
||||
found = f->current_offset != cp;
|
||||
|
||||
np = 0;
|
||||
LIST_FOREACH(matches, m, j->matches) {
|
||||
uint64_t q, k;
|
||||
|
||||
for (k = 0; k < n; k++)
|
||||
if (c->entry.items[k].hash == m->le_hash)
|
||||
break;
|
||||
|
||||
if (k >= n) {
|
||||
/* Hmm, didn't find any field that matched, so ignore
|
||||
* this match. Go on with next match */
|
||||
|
||||
found = false;
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Hmm, so, this field matched, let's remember
|
||||
* where we'd have to try next, in case the other
|
||||
* matches are not OK */
|
||||
q = le64toh(c->entry.items[k].next_entry_offset);
|
||||
if (q > np)
|
||||
np = q;
|
||||
}
|
||||
|
||||
/* Did this entry match against all matches? */
|
||||
if (found) {
|
||||
*o = c;
|
||||
*p = cp;
|
||||
return 1;
|
||||
}
|
||||
|
||||
/* Did we find a subsequent entry? */
|
||||
if (np == 0)
|
||||
return 0;
|
||||
|
||||
/* Hmm, ok, this entry only matched partially, so
|
||||
* let's try another one */
|
||||
cp = np;
|
||||
}
|
||||
}
|
||||
|
||||
int sd_journal_next(sd_journal *j) {
|
||||
JournalFile *f, *new_current = NULL;
|
||||
Iterator i;
|
||||
|
@ -168,14 +272,7 @@ int sd_journal_next(sd_journal *j) {
|
|||
Object *o;
|
||||
uint64_t p;
|
||||
|
||||
if (f->current_offset > 0) {
|
||||
r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
|
||||
if (r < 0)
|
||||
return r;
|
||||
} else
|
||||
o = NULL;
|
||||
|
||||
r = journal_file_next_entry(f, o, &o, &p);
|
||||
r = move_to_next_with_matches(j, f, &o, &p);
|
||||
if (r < 0)
|
||||
return r;
|
||||
else if (r == 0)
|
||||
|
@ -203,14 +300,7 @@ int sd_journal_next(sd_journal *j) {
|
|||
if (j->current_file == f)
|
||||
continue;
|
||||
|
||||
if (f->current_offset > 0) {
|
||||
r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
|
||||
if (r < 0)
|
||||
return r;
|
||||
} else
|
||||
o = NULL;
|
||||
|
||||
r = journal_file_next_entry(f, o, &o, &p);
|
||||
r = move_to_next_with_matches(j, f, &o, &p);
|
||||
if (r < 0)
|
||||
return r;
|
||||
else if (r == 0)
|
||||
|
@ -532,7 +622,7 @@ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret) {
|
|||
if (f->current_offset <= 0)
|
||||
return 0;
|
||||
|
||||
r = sd_id128_get_machine(&id);
|
||||
r = sd_id128_get_boot(&id);
|
||||
if (r < 0)
|
||||
return r;
|
||||
|
||||
|
@ -578,14 +668,18 @@ int sd_journal_get_field(sd_journal *j, const char *field, const void **data, si
|
|||
|
||||
n = journal_file_entry_n_items(o);
|
||||
for (i = 0; i < n; i++) {
|
||||
uint64_t p, l;
|
||||
uint64_t p, l, h;
|
||||
size_t t;
|
||||
|
||||
p = le64toh(o->entry.items[i].object_offset);
|
||||
h = o->entry.items[j->current_field].hash;
|
||||
r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
|
||||
if (r < 0)
|
||||
return r;
|
||||
|
||||
if (h != o->data.hash)
|
||||
return -EBADMSG;
|
||||
|
||||
l = le64toh(o->object.size) - offsetof(Object, data.payload);
|
||||
|
||||
if (l >= field_length+1 &&
|
||||
|
@ -613,7 +707,7 @@ int sd_journal_get_field(sd_journal *j, const char *field, const void **data, si
|
|||
|
||||
int sd_journal_iterate_fields(sd_journal *j, const void **data, size_t *size) {
|
||||
JournalFile *f;
|
||||
uint64_t p, l, n;
|
||||
uint64_t p, l, n, h;
|
||||
size_t t;
|
||||
int r;
|
||||
Object *o;
|
||||
|
@ -638,10 +732,14 @@ int sd_journal_iterate_fields(sd_journal *j, const void **data, size_t *size) {
|
|||
return 0;
|
||||
|
||||
p = le64toh(o->entry.items[j->current_field].object_offset);
|
||||
h = o->entry.items[j->current_field].hash;
|
||||
r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
|
||||
if (r < 0)
|
||||
return r;
|
||||
|
||||
if (h != o->data.hash)
|
||||
return -EBADMSG;
|
||||
|
||||
l = le64toh(o->object.size) - offsetof(Object, data.payload);
|
||||
t = (size_t) l;
|
||||
|
||||
|
|
Loading…
Reference in a new issue